Selection of Prediction Factors of Gross Primary Productivity Based on Artificial Neural Network

The analysis of drivers of gross primary productivity (GPP) in evergreen coniferous forests is an important part of forest carbon flux research. This study developed an artificial neural network (ANN model for coniferous forest GPP simulation based on the 1941599 half-hourly records of the flux observation network. The correlation coefficient(R) of observed and the predicted values ANN model base on all factors was between 0.89-0.94, and that of the combination of four factors was between 0.89-0.9. The experimental results illustrate that the method can mine the main driving factors for predictive GPP excellently without loss of precision. ANN model base on four factors made a good prediction of GPP. The climatic factors, which were dominantly contributing the prediction of GPP, include net radiation (NETRAD), vapor pressure deficit, air temperature, and soil water content/ carbon dioxide concentration.